Summary:
During shutdown we could fail to schedule callback in thread pool, in this case callback is invoked in the caller thread (reactor thread).

There are some handy classes, that was designed to be used from test, so they don't care about various complex things like shutdown.
Unfortunately CDC/xCluster code actively use code designed for tests.
As result in tries to fetch data during shutdown until timeout expires.

So patched this logic in outbound call to invoke callback with schedule failure status.
It results in aborting of retryable RPC calls such as read, write etc.
So particular tests complete 1-2 minutes faster.
Jira: DB-14444

Test Plan: Jenkins

Reviewers: hsunder

Reviewed By: hsunder

Subscribers: ybase

Tags: #jenkins-ready

Differential Revision: https://phorge.dev.yugabyte.com/D40612